Deep Matrix Factorization Model Combined With User Reviews

With the rapid development of Internet technology,the problem of information overload is becoming more and more serious in the era of big data.A recommendation system that can provide personalized information services has become the main technology to solve the above problems,and the matrix factorization model among them has been widely concerned because of its simplicity,efficiency,and high accuracy.The traditional matrix factorization model and other collaborative filtering methods that only use user ratings information on products are easily affected by sparse ratings and have poor results.In real life,when a user buys a product,he will refer to other users 'comments on the product.The user reviews that are widely available on the Internet can reflect the user 's preferences and the characteristics of the product.The research of algorithm provides an effective solution to the problem of sparse scoring matrix.Inspired by the idea of Bias SVD,this thesis uses the network framework of the deep matrix factorization model,combined with user ratings and comment information,to propose a new deep matrix factorization model DMFCUR.The model uses a multi-layer feed-forward neural network to learn user preferences and product features from the scoring matrix,and uses a convolutional neural network with Attention mechanism to learn user bias and product bias from the review text to obtain the latent factors of users and products.And then use the product of the latent factors of users and products as the predicted value of the score to achieve personalized recommendation.At the same time,this thesis gives a formal explanation that the DMFCUR model is a neural network implementation of Bias SVD,which provides a theoretical basis for its effectiveness.Finally,a lot of experiments have been carried out on real data sets,and the results show that the model proposed in this thesis performs better than the existing mainstream methods on very sparse data sets.